Hopf bifurcation analysis of some second order FDEs and neural networks with infinite delay (Q1355809)

This text deals with a general form of a second-order functional differential equation with infinite delay. Sufficient conditions are given, verified by the equations parameters, which assure the existence of bifurcations, giving rise to a periodic solution from an equilibrium point. These bifurcations are such that all the roots of the associated characteristic equation have negative real parts except at least a pair of purely imaginary roots. The paper concerns the case of only one pair of purely imaginary roots. With respect to a previous publication, related to the case of a finite discrete delay, this contribution provides a complete analysis of a more complex situation of an infinite delay. This is characterized by the presence of an integral of the delayed variable having one of the two limits which is infinite. Two examples illustrate the sufficient conditions. The first one concerns a generalization of the normalized Hopfield's model of neural networks when a distributed delay is present. This application is not limited by that (two) of the theory. The second illustration is related to a predator-prey model with saturating predator. According to a bad habit picked up since a long time, the bifurcation considered here is attributed wrongly to Hopf. It is recalled that in his paper (1942) Hopf mentions that he deals analytically with a bifurcation described by Poincaré, and that the first analytical study is due to Andronov (1938).